Data Ownership Rights and Controls October 2018

Data ownership,
rights and controls:

Reaching a common
understanding
Discussions at a British Academy,
Royal Society and techUK seminar
on 3 October 2018
Data ownership, rights and controls: reaching a common understanding
Contents
1.0 Summary 2
2.0 Introduction 3
3.0 Data ownership, rights and controls: discussions at a

British Academy, Royal Society and techUK seminar on
3 October 2018 4
3.1 Exploring the concepts 4

3.2 Moving the debate forward 7
3.3 Getting to where we want to be 8
4.0 Contributor papers 10
4.1 Legal notions of ‘property’ and ‘ownership’ -

Professor Sarah Worthington FBA 10
4.2 Data ownership: is it an appropriate concept? -
Eleonora Harwich, Reform 16
4.3 Data: vital asset or toxic liability? -
Professor Jim Norton FREng 16
4.4 The HAT data ownership model: first party IPR for individuals -
Professor Irene C L Ng 18
4.5 Data ownership and data rights -
Roger Taylor, UK Centre for Data Ethics and Innovation 24
4.6 Reflections on ‘data ownership’ -
Guy Cohen, Privitar 26
4.7 Reflections on the data ownership, rights and controls seminar
from the ICO -
Romin Partovia, Information Commissioner’s Office 24
4.8 Data ownership, rights and control – an anthropological
perspective -
Dr Hannah Knox 26
5.0 List of speakers 28
1
1.0 Summary
On 3 October the British Academy, techUK and the Royal Society convened a seminar,
which provided an opportunity to explore and understand the concept, value and
limitations of the idea of ‘data ownership’. It considered the sound bases from which to
consider and probe the concept of data ownership and discussed issues relating to the
ability to exert rights and control over data use.
Based on the discussion during the seminar the following key points have been
identified as warranting further consideration and discussion moving forward:
• Use of the term “data ownership” raises significant challenges and may be
unsuitable because data is not like property and other goods that can be
owned or exchanged
• Instead discussion should explore the rights and controls individuals, groups
and organisations have over data, and should encompass a societal as well as
individual point of view
• Broader debate could help to better describe the data rights and controls that
are often associated with the concept of ‘data ownership’.
This paper summarises the rich and diverse discussion at the seminar, and is followed
by a set of papers, which provide further explorations of data ownership, rights and
controls.
Disclaimer
This is a note summarising the discussion and debate at the British Academy, Royal
Society and techUK event on Data ownership, rights and controls: reaching a common
understanding. It is not intended to represent the views of the British Academy, the
Royal Society or techUK, nor does it represent the views of individual attendees of the
event. The ideas and reflections contained within are not necessarily endorsed by the
British Academy, Royal Society or techUK.
2
2.0 Introduction
In 2017, the British Academy and the Royal Society published Data management and
use: Governance in the 21st Century. This report addressed a changing data landscape
and recommended the need for a new governance framework for data use, based on
the principle of human flourishing, and with a need for a new body to steward the
landscape as a whole.
In the 18 months since the publication of this report, there have been significant
changes in the data governance landscape. The General Data Protection Regulation
(GDPR) came into force in May 2018, requiring data protection ‘by design and by
default’. 2018 saw the UK government establish the Centre for Data Ethics and
Innovation, complete a consultation on the Centre’s role and activities, reflecting many
of the recommendations made in Data management and use, and appoint the Board
members that will help to steer the work of the Centre going forward. Also in the last
year, the Nuffield Foundation established the Ada Lovelace Institute, which has a
mission to ensure data and AI work for people and society.
On 3 October 2018 the British Academy, the Royal Society and techUK held a seminar
to reflect on these changes, and to explore a key issue within the broader conversation
about data management and use: data ownership, rights and controls.
The Data management and use report argued that some of the core concepts that
underpin governance of data use are challenged by current technologies and data
management practices. These include consent and the idea of data ownership. The
report states: ‘Uncertainties around the concept of ownership can be a barrier to
effective trade and transfer of data, and leave individuals and organisations uncertain
about their rights.’
Ensuring that individuals and organisations are able to exercise the appropriate rights
and controls over data is essential to the data economy and is at risk unless we build
the right approach to data rights and control.
The seminar was therefore held to provide the opportunity to explore and understand
what is meant when individuals and groups refer to ‘owning’ data or ‘my’ data, and to
explore the concept, value and limitations of data ownership from individual and
organisational perspectives, in both the private and public sectors. It considered the
sound bases from which to consider and probe the concept of data ownership and
discussed issues relating to the ability to exert rights and control over data use and
assessing and accessing the value of data.
This report includes reflections of the discussion at the seminar, and a set of
contributed papers. These include papers submitted ahead of the seminar to stimulate
discussion and papers submitted after the seminar to expand and open up areas for
further discussion.
3
3.0 Data ownership, rights

and controls: discussions
at a British Academy,
Royal Society and techUK
seminar on 3 October
2018
Where are we today? Data collection has grown massively, increasingly encroaching on
private spaces. We have become very used to giving up our data, consenting to its
gathering and to its use, often in ways we dimly appreciate, in exchange for ‘free’
access to products and services that we value. This can create a feeling of unease,
and this feeling is amplified by high-profile stories about the mismanagement and
misuse of personal data. 1
This sense of unease is a difficult problem to pin down, especially as it is in tension

with an awareness that data can do much good – for example in its use in health
research. The data governance regime faces the challenge of recognising the
enormous potential for public good, but also the potential for both very specific and
general harm. It also has to deal with both individual data and the holders of large data
sets, and it has to balance the interests of the individual and society.
The systems for governing data are clearly under stress along with the concepts we use
to talk about data. Data ownership is one of these rather problematic concepts – what
does it really mean? Can you own data? Data is replicable and is not something that
you use up, potentially you can share it as much as you like. If you can own data, under
what circumstances and who should be able to own it? Are there different
considerations in the private and public sectors? What does it mean for individuals and
their ownership of data about them when the value of data comes from a collective
dataset rather than from the data about one person?
The seminar highlighted that the debate and discussion around data ownership, while
nascent, is becoming more and more significant given the increasingly data-enabled
society in which we all live and work.
3.1 Exploring the concepts

The uncertainties that surround concepts like ‘data ownership’ and ‘data rights’ have
created significant barriers to debates on data. Which aspects of these concepts are
important, and which need to be revised?
3.1.1 Data ownership
The concept of ‘data ownership’ seems to have quite a lot of intuitive power. ‘Your data’
seems to be a simple shorthand for data that is about you, and because we feel as
though we understand how ownership works, this seems to be a helpful way to get
purchase on ideas that are otherwise difficult to talk about. Motivations for talking
1 , Including the Facebook-Cambridge Analytica scandal in March 2018.
4
about ownership include privacy protection, the desire to be able to use one’s own data
(both for individuals and organisations), and the idea of sharing in the benefits that
others get from using data that might be about you as a person.
It seems intuitively right that you should have control over ‘your data’, and that if it were
used for financial (or even political) gain that you should be able to benefit. This was
seen clearly in many of the reactions to recent high profile incidents in which many
people were upset by how ‘their data’ had been used for political purposes without
their knowledge or consent.
However, there are very significant problems with the concept of ‘data ownership’ that
make it unsuitable for use in developing a vision for a system of data management that
combats the growing sense of unease.
The idea of owning data is challenging because data is not like other goods that we can
own. It is non-rivalrous – I can both give it to you and still have it myself without it
costing me any of the original good. Other goods are not like this. If my bag is stolen, I
no longer have it. But, generally, if your data is stolen you still have it, but someone else
has it too. If I sell my house to you, it is yours, it no longer belongs to me and I cannot
sell it to someone else, but this is not always the case with data, be it personal data or
data that is not about people at all.
In addition, data can be about multiple people, breaking the link between the idea of
data that is ‘about me’ and data that I therefore ‘own’. For example, genetic data about
me is also about my family, and data that is produced through a business or other
relationship with other people or organisations also inherently involves other people.
Data about your purchases and preferences is often also about friends and
acquaintances. Conversely, personal data about individuals retains a connection to an
individual, it is still ‘about’ them, even if it has in some way been transferred to
someone else. It is also not clear why the subject of the data would be the data ‘owner’.
The parallels to other forms of property are actually easier to see if the person
understood to ‘own’ the data is someone who holds an aggregated data set about
many people.
For these reasons, there is a lack of legal basis, in common or civil law, for the idea of
data ownership. Common and civil law lack a definition of ‘data’ and do not confer a
special status on it. Only personal data is defined, non-personal data is not defined,
and even with personal data there is no clarity whether it can be held or not, and the
definition of ‘personal data’ is extremely broad. It is also a dynamic concept: what is
today not personal data could be considered in the near future to be personal data if
changes mean that it can be used to identify an individual. Technology evolves
continuously and even machine-generated data could be considered, in some
situations, as personal data.
Anthropology considers ownership in relation to the social practice of exchange. It is

primarily at the moment of exchange, when one person gives something to another
person, that the very question of ownership is made visible. One of the things that is at
stake in debates about data ownership might be not only data’s (lack of) legal status as
property, but also its social status as an artefact of exchange. Could some of the
problems about what constitutes appropriate exchange in fact be what is at the heart,
in some of the discussions, about data ownership?
3.1.2 Consent and control
A different approach is to place the focus on consent for data use in order to give
individuals a sense of control. This shifts debate from ownership to the control that
people should be able to have over the data that is about them.
5
This approach has the advantage of directly addressing a concern that has been at the
heart of data management scandals: that data about individuals is used without their
informed consent.
It was raised in the discussion at the seminar that most legal protection of individuals
can be waived by means of consent, which creates a risk as there are two significant
problems with both this approach and with the concept of ‘data ownership’ discussed
above. The first is that both approaches can place an unreasonable burden on
individuals. Individuals may not have the ability, knowledge or time to make informed
decisions about all the uses of data about them twenty-four hours of the day. It may be
that we make imperfect decisions when we give consent, and this is not just about
laziness, it may also be about the complexity of understanding what the implications
are of giving consent for particular data collection and use.
The second problem is that there is a risk to presuming that individuals are always able
to give consent. The reality is that the data processing is integral to the delivery of many
services and processes and data can be processed legally under other legal
mechanisms such as legitimate rights. If individuals were genuinely expected to give
permission for every use of data, they would spend their lives doing nothing else. This
can in turn create unrealistic expectations of the amount of control individuals might be
able to exercise over data about them, increasing the sense of unease when
expectations are unavoidably unmet.
Further, when individuals say that they would like to exercise control over data about
them, some might argue that they do not, in practice, take advantage of the control that
they already have. For example, individuals may rarely read terms and conditions
carefully before clicking ‘I consent’, because the terms and conditions are lengthy, and
seen as hard to understand, yet they are the gateway into accessing a service or
system. The fact that people may, for many understandable reasons, fail to make
genuine use of opportunities to provide or withhold consent has the effect that we
rarely think in detail about all the different ways in which data is being gathered,
aggregated and used to make our lives better. Instead, individuals’ interest in consent
and control is primarily spiked only when something goes wrong.
3.1.3 Data rights
One way to tackle some of the limitations of an approach based on data ownership
would be to think instead about a conceptual foundation of a bundle of data rights. This
approach holds that individuals have rights over data that is about them. This does not
necessarily require that they have control over data in the same ways that individuals
have control over property that they own. Rather, individuals might have a right to data
about them being used in only fair and reasonable ways, or a right to have personal
information anonymised in any aggregation that is publicly available.
This approach may have significant power because it builds on a sophisticated and
nuanced body of thought that is well-suited to considering fairness, balance and trade-
offs. Furthermore, rights are generally well-understood, which might go some way
towards addressing the difficulties in communication on data issues.
Additionally, a rights-based approach helps to ensure at least a minimum degree of

equality, which in turn underpins the need to ensure that the system of data
management is fair.
A further benefit of taking this approach is that it requires careful thought about
corollary duties. This is helpful as a way of linking individuals’ rights over data about
6
themselves to the bodies that carry out stewardship functions. These bodies might hold
the related duties, or at least carry out the duties on behalf of the state.
However, existing conceptions of human rights are highly individualised in their

orientation and were not devised with data in mind. Hence, taking data rights as the
underpinning conceptual framework requires further careful thought to answer many
outstanding questions. For example:
• What is the relationship between individual data rights and collective data rights?
• Which data rights are important, and which should take priority?
• Which bodies hold the corollary duties?
• How are data rights changed by processes of anonymisation?
3.2 Moving the debate forward

An alternative approach to thinking about ownership, consent, control and rights is to
place the focus on trust in the data management system. Instead of placing the burden
on individuals to make decisions about data about them, we can consider what sorts of
institution, accountability and regulation will give individuals trust in the system? One
way to approach this is to consider the role of a fair, trustworthy steward. This approach
presumes that individuals would rather feel confident that data about them is being
looked after than have to make decisions about data themselves. For this approach to
work, individuals need to trust the data management system and its underpinning
infrastructure. This requires institutions and regulations that protect individuals, and
that are accountable for their work.
This approach leaves space for considerable nuance and an understanding of the
trade-offs inherent in the use of data. There are a number of ways we can move
forwards in this direction.
3.2.1 Taking a societal view
A focus on data ownership tends to be individualistic. But does society have rights over
data that is about us? The census, for example, is an important activity that benefits
everybody and necessarily infringes on some of our rights over data about us, as does
for example surveillance that enables public safety. These are things that protect us or
benefit us that we have to balance against our individual rights.
Taking a societal view also highlights some of the shared risks of our data-enabled
society. We are already in a world where there are the data-rich and data-poor. We can
all benefit when those who are in power have access to high-quality data, because it
can enable them to make better decisions. But we have to equip those who do not
currently have the capability, whether that is through resources or training, in order to
try to rebalance power so that everyone can use data for their own good.
3.2.2 Understanding the value of data
From an anthropological point of view ownership and exchange are key concepts for
understanding societies. We need to better understand what happens when data is
exchanged, and the value that is involved in an exchange of data – what we get back
and what we give away. This is not only about what you might get in exchange for data
immediately or in the next year, but also about how data, once processed, may have an
effect on your life in the future.
A challenge is that there is a huge mismatch in size between individuals and the
organisations with which they exchange data. How does the individual negotiate in
7
those contexts? Better understanding of data value and how to exchange on fair terms
is essential.
3.2.3 Focus on use
Both the value and the harm from data come from purpose for which and the way in
which it is used. We have tended to focus on controlling the collection of data, but
there should be a shift in focus toward the use of data and the impact of that use on
individuals. Focusing on what we want at the point of data use may be more effective
in reducing harms than focusing on data collection, although that too remains
important. While there is no clear law on owning data, there are laws to stop people
doing ‘bad things’ with data. Focusing how we control what people can do with data
may be more valuable than trying to establish how we protect data as property. In
addition, we need to convey that there are significant social benefits to making use of
data and moving away from debates on ownership can help to open up debates on how
to create a system that uses data for public good but minimises the negative
consequences.
3.3 Getting to where we want to be

3.3.1 Building trust
An approach that focuses on building a trustworthy system and providing stewardship

requires institutions and regulations that protect individuals, and that are accountable
for their work. A significant benefit of taking this approach is that many of these
institutions and regulations already exist, such as the Information Commissioner’s
Office (ICO) and GDPR, and place an explicit focus on principles such as transparency
and fairness, which are likely to inspire trust.
3.3.2 Engaging the public and building a social vision
We do not have a sufficiently deep understanding of public attitudes with regard to

their relationship with those who are using their data. But in order to find out, we must
first acknowledge the complexity of the concepts in play. We can engage with the public
on questions about concepts by focusing on the information that citizens need to make
the choices they think they want to make, and by asking about the sorts of institutions,
accountability and regulation that will build trust in the system. This goes beyond
questions of data into wider questions about the kind of society that we want to create.
This requires that the individuals and institutions that make up the current data
management system get out of their ‘bubble’. This might include:
• Engaging specific sectors and looking at specific use cases to develop the detail of
how this approach might work in practice, identifying challenges and adjusting in
response.
• Ensuring a wide debate across society, across all parts of the country.
• Working closely with civil society organisations to understand what is causing the
sense of unease and how best to combat it.
• Considering how to engage with groups of a wide range of sizes, from small civil
society groups and charities to large private sector companies and government
departments.
Taking this approach also requires serious thought about a wide range of questions for
further consideration. These include the questions in section 2.1 (3) that aim to further
flesh out the concept of ‘data rights’, which might helpfully underpin a data
management system focused on stewardship. Further questions include:
8
• How can the system be built in a way that is ‘future proof’ and adaptable, given the
speed of technological change?
• How can equality and trustworthiness be best ensured in a system that includes
extremely large and powerful profit-making organisations?
• How does this framework for thinking about data management relate to broader
questions about the kind of society that we wish to create?
• What does ‘stewardship’ look like in different sectors?
3.3.3 Conclusion
With a number of existing institutions exploring ethical issues in relation to data and
advanced digital technologies including bodies newly created since the publication of
Data management and use, we need a common foundation on which to build debate.
This debate should move on from ownership to how we understand and manage the
balance between collective and individual benefits, risks, rights, and the balance
between the interests of individuals, groups and industry. If there is a concern with
ownership, it might be that what people would really wish to achieve is that those
balances are somehow fair, that there is a balance between who is owning or
experiencing the risk, and who is owning the value from data and experiencing the
benefit that come from data use. There may be a role for an overarching framework
that addresses these issues, but exact resolution is likely to be different in different
cases. There is no panacea through a concept of data ownership, but there is a need
for a set of specific discussions on how we use data. Current institutional change and
new bodies mean that we are now building the capability and capacity needed to
enable these discussions to happen.
9
4.0 Contributor papers

The following set of contributor papers were submitted ahead of the seminar to
stimulate discussion and after the seminar to open up and expand the discussion. The
ideas and reflections contained within are these papers are not necessarily endorsed
by the British Academy, Royal Society or techUK.
4.1 Legal notions of ‘property’ and ‘ownership’

Professor Sarah Worthington FBA
This paper is adapted from Professor Worthington’s keynote address at the seminar.
The policy and governance dilemmas associated with data rights and data control are
intensely challenging. Ideas of property ownership and control – or ‘data ownership’
and ‘data control’ – are increasingly seen as providing possible solutions to these
complex issues.
As a non-expert in this field, but as a lawyer whose special subject is property, I want to
make some general comments about the legal notion of property. My aim is to provoke
and focus discussion amongst the experts who are deeply engaged in the detail of what
these concepts mean for data.
I make four main points. First, it is essential to be very clear about the end goal before
selecting the appropriate legal means of getting there. Once in place, a legal rule will
have consequences. It is important to ensure these are the intended consequences,
not unintended ones. Secondly, having ‘property’ is not as protective as one might
think. This may have important consequences in thinking about data rights and data
control. Thirdly, and on the plus side, even though ‘property’ may not be quite as
protective as often assumed, English property law is nevertheless rather remarkable,
and worth a little investigation. Fourthly, the obvious alternatives to ‘property’ thinking
in data governance may be worth deeper investigation.
4.1.1 What is the regulatory end goal?
Defining the end goal is typically the hardest part of any project. Here is no exception.
If we want the right answers, then we have to make sure we ask the right questions.
In discussions about data, focus typically centres on the personal information we hand
over to third parties and the use third parties make of that information, in particular the
use they make of aggregated data sets or the resulting ‘data infrastructure’.
Until recently there has been relatively little public concern about individual data
collection. The law does not ban the mere observation of individuals going about their
daily business. This means that publicly observable shopping habits, movie watching
habits, newspaper reading habits, height and weight estimations, etc, might all be
gathered without restraint (although use of the gathered data might be more
constrained). But now this data collection has encroached on ‘private’ places:
machines – not people – log all the above habits and more besides, whether these
activities are carried out in public or at home. Moreover, these diverse data points can
all be integrated, again by machines, aggregating face recognition data, credit card
usage data, mobile phone location data, and the list goes on. This integration can be
done almost instantaneously, when previously the aggregation might have taken
months or years to produce by private investigators focusing on one individual at a
time. That possibility feels invasive in and of itself.
10
Modern data usage goes still further, with each of us being increasingly complicit in it.
We bargain – or ‘barter’, as Gillian Tett described it in last weekend’s FT (6 October
2018) – with our own data records, giving up the data or consenting to its gathering
and its use, often in ways we dimly appreciate, in exchange for free access to products
we value.
In any event, our own individual data point is, by itself, of very little value (either to us or
to the data collector), so we perhaps quite rightly feel as though we are obtaining the
online services for free. Yet the service providers freely acknowledge that it is the data
we deliver that is valued, not any funds we pay over (FT, 13 October 2018 with the
focus on Monzo).
Only very recently have we begun to understand that this may be a pact with the devil.
We are now increasingly likely to be offered on line for sale only what we have already
indicated we wanted to buy, or the types of movies we once enjoyed, or indeed the
news and political views that we once chased down. And from there it is a very short
and slippery slope to the ‘infrastructure of industrialised persuasion’ that Onora O’Neill
rightly flags as a public risk to our political and social institutions, not just a private risk
to individual data subjects.
And yet, dramatically, on the other side of the leger, the public benefits of data
collection and its use are legendary. A great deal of both ancient and modern medical
research is based on the analysis of data sets. The same is true of old and new social
policy interventions. Think of Charles Booth’s poverty maps of London, or even the
Doomsday Book, alongside their modern equivalents.
No doubt we want all the good and none of the bad. It is always so. In the industrial
revolution people wanted the jobs and cheap goods, but not the pollution and slave
wage push.
Here the hard question is what we will put up with, or even welcome, and what we will
bar. This is difficult in the extreme, and yet this question is one that must have a clear
answer before we can have any hope of designing a data governance regime or a legal
infrastructure that will deliver the desired ends.
4.1.2 ‘Property’ is not as protective as one might think
Because property is viewed as being especially protected by the law, concepts of data
ownership seem to be the answer to any difficult question in this area. However,
property is not as protected as most people think.
Take a simple illustration. I own my bicycle. Most people would expect that if it were
stolen the law would ensure that I could get back. However, English law holds to the line
that even if I can find the thief I am only entitled to money, not to the bicycle. Most non-
lawyers find that response startling and completely counterintuitive: what is the point of
owning things if you cannot even recover them from a thief?
Predictably, a statute creates exceptions (the Torts (Interference with Goods) Act 1977
s. 3). If it is not my bicycle that has been stolen, but a Picasso painting, then I will
indeed be able to recover the painting. You might think that personal data is more like
the Picasso painting than the bicycle, but there are further difficulties with data that I
shall come to shortly.
It is also important to note that rights, entitlements and the power to control need not
necessarily be associated with ownership. The right to light is not associated with
ownership of light, and control over the export of national art treasures does not
11
indicate that the government owns all these art treasures. If rights and entitlements
are identified as valuable, they can be allocated and protected without any
intermediating notion of ‘property ownership’. That simple fact ought not to be lost in
discussions.
Notwithstanding this simple fact, ‘ownership’ and ‘control’ are increasingly prevalent in
discussions about regulating data usage. If this is where the focus is to lie, then it is
crucial to understand something about legal notions of ‘property’. The law’s approach
to property is not necessarily aligned with the general public’s assumptions.
4.1.3 How do lawyers think about ‘property’?
English law, and all the common law systems derived from English law, have long had
an approach to ideas of ‘property’ that is quite different from civilian jurisdictions.
Nowadays these latter jurisdictions increasingly seek to mimic by statute what the
English courts created independently centuries ago.
The first thing to notice is that English law, unlike its civilian counterparts, no longer
holds to a sharp divide between tangible assets (such as land, bicycles and Picasso
paintings) and intangible assets (such as shares, bonds and debts). Put in legal terms,
English law no longer draws hard lines between ‘property’ and ‘obligation’ or ‘property’
and ‘contract’. All these different types of rights are ‘assets’: they are all different forms
of wealth. All are valuable, all can be controlled in similar ways, all are protected by the
law in similar ways, all can be used in commercial transactions in similar ways, and so
on. Recall all the commercial assets in issue in the last financial crisis. They were not
tangible ‘things’; they were merely contractual rights, yet they could be traded
internationally, secured, made the subject matter of trusts, and so on.
If that is true, then what makes ‘property’ special in English law? The answer, it now
seems to me, is simple but rather surprising. It is an idea I only settled on relatively
recently, but it seems to have a good deal of explanatory force. English property law is
not about classifying assets as ‘property’ or ‘not property’, with some differential
protection accorded to each class. Rather, it is all about the sharing of assets,
whatever their type.
We are very used to this idea in the context of tangible assets. The ability to split legal
ownership from possession enables legal owners to lease land, hire out a car, pledge a
Picasso painting as security for a loan, and so on. All these options are possible simply
because we – along with every other legal regime no matter how primitive – recognise
the commercial and social advantages of being able to split ownership from possession
and share an asset in various ways, retaining its ownership in one person but
permitting someone else to have possession on specified terms.
Notice that such an arrangement is necessarily a transactional arrangement between

the two parties concerned in the sharing arrangement. The terms of the sharing have
to be agreed, whether the arrangement is a contract of hire, or a gratuitous loan, or a
pledge to secure a loan. And if it is possession that is being granted, then the subject
matter – the asset – has to be a tangible asset that can indeed be possessed.
English law also recognises other forms of sharing that are less obvious. For example, it
recognises various forms of security interests, which can be taken over tangibles, such
as houses and machinery, but also over debts, or shared, or indeed entire businesses.
And if that were not enough, English law has gone still further and allowed the creation
of trusts. These too are possible over any type of asset at all. The trust enables a legal
owner to specify that certain economic benefits that might be derived from a
designated asset will be held for some nominated third party (the beneficiary of the
trust) rather than for the owner. It is possible to have trusts of historic country estates
12
and grand family art collections, or trusts of shares held for nominees, pension trusts,
superannuation trusts, or client account monies or investment funds held on trust. The
term ‘trust’ simply signals that although one individual is the legal owner of the asset,
the nominated economic benefits that might be derived -from the asset are held for
other parties, not for the owner – another form of sharing.
To summarise, ‘property’ in English law is not some special classificatory system that
divides assets into ‘property’ and ‘not-property’ classes and protects each class
differently from the other. Instead, ‘property’ in English law is a set of rules that
enables legal owners to share the benefits of their assets with third parties by way of
different types of derivative interests, whether those derivative interests are
possession, security interests or trust interests. These legally enabled sharing regimes
have proved invaluable, both socially and economically.
4.1.4 Data as property?
But notice one crucial limitation. In all of this it is essential that there is ‘an asset’ of
which one might say ‘I own this asset’ or ‘I am determined to share this asset with X’.
And it is here that the crunch point arises in relation to any discussion of ‘data
ownership’ and ‘property in data’. In most legal systems, information, or ‘data’, is not
an asset.
Why is that? For a start, with information there is not the same ability to control
access, or assignment, or – crucially it seems – sharing. If you steal my bicycle, I may
still have legal title but I no longer have possession. It is very clear what I have lost and
what you have gained by your criminally enforced sharing with me. By contrast, if you
steal information, we both have it. We then need to think very carefully about what I
have ‘lost’ by way of involuntary sharing that the law should remedy.
In some areas the law has taken up with a vengeance this challenge of dealing with
information and ideas. This is where intellectual property has a crucial role to play. All
intellectual property rights are created by statute, not by the courts. Notably, despite
the ‘property’ terminology, the protection delivered by these statutory means is not
dependent on any idea of there being ‘property’ in the creative idea or endeavour.
Instead, the statute itself defines rights, and then defines remedies for their
infringement, and it is these statutory rights that are then ‘assets’ that may be
assigned or shared in all the ways that other assets can be dealt with at law.
In creating these statutory rights, the relevant statute defines the scope and extent of
the rights in issue (note how carefully that is done) and then provides for time-limited
monopolies over those rights to the creator or inventor. The commercial privilege of a
limited monopoly, a monopoly that can itself be sold or shared by licence, is given by
the law in exchange for full public access to the rights and their inherent creative and
knowledge benefits in the longer term. But in the interim, before public release, the
statute provides the creator with remedies against those using these statutory
protected benefits without consent.
As these rights are currently structured, they do not help very much with data
protection. The individual subject’s data points of the sort collected so commonly
these days do not fall within any existing statutory definition of protected ‘intellectual
property’. There is some limited protection afforded to data sets by way of database
rights, but this does not protect the individual subject of the data from the collection of
the data nor from its aggregation, but instead protects the holder of the aggregated
database from its use by competing third parties who might want access to the
compilation either in whole or in part.
13
We should not really be surprised that general data – the individual subject’s data
points – is left unprotected by this legislation. Recall the practical and policy
considerations raised earlier. If protective legislation were to be enacted, then it would
likely be very difficult to define what controls and what permissions we would want
embedded in the legislation, especially given the recognised public benefits of data
gathering and analysis, and notwithstanding that these benefits need to be balanced
against the potential for substantial personal and public detriments.
Perhaps more difficult still is that it would be quite hard to define what sort of initial
data collection would be constrained. As noted earlier, English law has long held to the
idea that there is no property, and indeed no right of any individual, in a ‘public
spectacle’. So passers-by are free to observe other individuals going about their daily
activities, including their shopping and movie watching and internet browsing, all
without interfering with the subject’s legal rights. All the more so if the observation and
recording is done with the data subject’s consent, as it now so often is. Then it is
irrelevant whether the observation is of public or private ‘spectacles’ or behaviours and
information.
If the law is to have anything to say about this sort of general data collection and use –
and there appears to be general agreement that it should – then it would seem to be
essential to define some new kind of right for the data subject that is protected in
specified ways. In doing that, I suggest it would be quite unnecessary to label any such
statutory protection as ‘property’ or ‘data ownership’, notwithstanding the attractions of
such terminology. The data subject would simply have a legally protected right, or –
more likely still – the data user would be subject to specified legal constraints in its
activities. The notion that this can be done is not at all difficult. What is difficult is the
precise settling of the desired limits and permissions. That fraught debate goes to my
first point, which must of necessity be the issue that is addressed first.
4.1.5 Alternative strategies beyond ‘data ownership’
There are any number of alternative strategies that might be used in regulating data
collection and usage. I cannot range as widely as one might, but I want to make two
comments, one about consent and one about privacy.
Most legal protections of individuals can be waived by consent. Increasingly we give

consent to the collection and use of our data, often without investigating the terms of
that consent or understanding the ends to which the data might be used. Indeed, the
data gatherers themselves may not comprehend the potential ends for which the data
might be used. The parallels with the last major financial crash and the trading in
derivative interests that no one really understood is plain. In line with a great deal of
consumer protection legislation, it may be worth considering whether these ‘take it or
leave it’ agreements with consumers in relation to their data collection and usage
should be subject to constraints on what terms can be taken to be agreed by a
consumer simply clicking ‘I have read and accept the terms’.
Secondly, privacy and its legal protection is a much more ephemeral concept than
property. Rights to privacy are enshrined in the European Convention on Human Rights
(Article 8 ECHR delivering ‘the right to respect for your family and private life, your home
and your correspondence’) and in the UK Human Rights Act. But the meaning of privacy
is not defined, and nor is its protection absolute – it is a qualified right requiring the
balancing of its protection against the need to protect other similarly important
personal rights. The practical analogies with data usage and its governance are clear,
and thus the legal means that have been used to address the basic concept of privacy
and then undertake the necessary balancing act in engaging with its protection may
14
hold some important lessons in thinking about protection of an individual’s data and its
use.
4.1.6 Conclusion
In summary, any data governance regime faces the difficult task of dealing with the
individual subject of the data and also the holder of the aggregated datasets,
recognising the enormous potential for public good, but also for specific and
generalised harm. Whether the ‘property model’ of data ownership and control
provides the best legal approach to the necessary governance regime is a question
discussed in detail in other sections of this report.
15
4.2 Data ownership: is it an appropriate concept?

Eleonora Harwich, Director of Research and Head of
Digital and Tech Innovation, Reform
The digital trails that individuals leave when they go about their daily lives give clues
about who they are, what they do, what they want and need. 2 This information is
valuable to both the public and private sector. 3 Data on how individuals interact with
public services can be used to deliver better services and outcomes for individuals. 4
Nevertheless, data is not always used with the principle of human flourishing in mind
and there have been notable instances of misuse both in the public and private sector. 5
These often spark off debates around the notion of ownership as individuals feel that
data about them was used for something they object to or have not been informed of.
Despite the legitimacy of this reaction, is data ownership a useful and appropriate
concept?
People tend to assume an implicit understanding of the concept of data ownership,

which means it is often not defined. However, this may be because the notion of data
ownership does not exist in legal terms. A scan through the Data Protection Bill or the
General Data Protection Regulation shows no mention of the concept. As highlighted in
the House of Lords report AI in the UK: ready, willing and able?, the notion of ownership
cannot be easily applied to data and is therefore not fit for purpose. 6 Data is legally
inert in itself. 7
Instead, the law speaks about the rights and duties that arise in relation to data. 8
These need to be upheld by data subjects (i.e. individual who is the subject of personal
data), data controllers (i.e. a person who, either alone or jointly or in common with
other persons, determines the purposes for which and the manner in which any
personal data are, or are to be, processed) and data processors (i.e. any person, other
than an employee of the data controller, who processes the data on behalf of the data
controller).
Data is produced through transactional relationships so it is difficult to ascribe it

property rights and treat it like any other asset. Data is created through an interaction -
an individual with their doctor or an individual and their Internet provider, for example.
In addition, in certain contexts data about an individual can be revealing about that
individual’s family, which means that data about a person is not always only data about
that person.
Despite this, there is still a vocal group, particularly in the distributed ledger technology
community, that upholds data ownership as moral imperative. Their definition of data
ownership can be boiled down to three core elements: the right to access, to control
2 Daniel “Dazza” Greenwood et al., ‘Reshaphing the Social Contract: The New Deal on Data’, in
Trust:: Data, A New Framework for Identity and Data Sharing, ed. Thomas Hardjono, David Shrier,
and Alex Pentland (Visionary Future, 2016),103.
3 Ibid.
4 Sarah Timmis, Heselwood Luke and Harwich Eleonora, Sharing the Benefits : How to use data
effectively in the public sector, (Reform, 2018).

5 Health and Social Care Committee, Oral Evidence - Memorandum of Understanding on Data-
Sharing between NHS Digital and the Home Office, 2018; Zoe Kleinman, ‘Cambridge Analytica:
The Story so Far’, BBC News, 21 March 2018
6 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, Report of
Session 2017–19 (House of Lords, 2018), 28.

7 Richard Kemp, Hinton, Paul and Garland, Paul, ‘Legal rights in data’, Computer Law and Security
Review, 27, 2001, 142.

8 Ibid.
16
and to distribute one’s data. 9 This would be akin to the data subject becoming the data
controller. Individuals could access data about them at any time; individuals would
have full control over how data about them is used if they do not agree with the terms
they could entirely remove their data; finally individuals would be able to decide with
whom to share their data with. 10 They are the distributors data about them.
This model has some attractive and desirable features such as giving people more
agency and control over what can be done to data about them. These features are
crucial for building a trustworthy data infrastructure. 11 Nevertheless, building this type
of system does throw up challenges of its own. Firstly, there is a certain amount of data
that is necessary to share for services to be able to work. For people not be incessantly
pestered with notifications about services wanting to access data about them, rules
would have to be put in place in a smart contract (i.e. a computer protocol that allows
for the transfer of a digital asset from one party to another automatically under agreed
upon stipulations and terms) about what is the minimum amount of data that should
be shared with different services. This already means giving up a bit of control. In
addition, deciding what is the minimum amount of data a service needs to function
might not always be a clear and objective decision. Secondly, who will build this system,
approve of its accuracy and fitness for purpose and who will carry out the oversight of
it? These are crucial question that do not seem to have a clear answer yet.
Individuals can have greater control over what happens to their data without the need
for ownership. The debate about data ownership is one, which reduces data to an
asset. This might be the way that many companies currently view data, but it might not
be the most practical or desirable definition. Some have argued that data should
instead be viewed as a form of “digital labour”. 12 This might better reflect the complex
nature of data.
4.3 Data: vital asset or toxic liability?

Professor Jim Norton FREng
4.3.1 Context
The Royal Society, in its previous two reports and broader programme of work on Data
Analytics and AI, has amply documented the vital role of access to data sets of
appropriate scope and quality as a lubricant of the 21st century economy. There remain
however unanswered questions on many aspects of data collection, ownership and
exploitation. This short provocation seeks to highlight areas where good progress is
being made and those worthy of much broader public debate.
9 Greenwood et al., ‘Reshaphing the Social Contract: The New Deal on Data’, in Trust:: Data, A New
Framework for Identity and Data
Sharing, ed. Thomas Hardjono, David Shrier, and Alex Pentland (Visionary Future, 2016),106.
10 Ibid; Guy Zyskind, Nathan Oz, and Alex Pentland, ‘Decentralizing Privacy: Using Blockchain to
Protect Personal Data’, in 2015 IEEE Security and Privacy Workshops, 2015, 180–84.
11 British Academy and Royal Society, Data Management and Use: Governance in the 21st Century,
2017; Timmis, Heselwood and Harwich, Sharing the Benefits : How to use data effectively in the
public sector, (Reform, 2018).
12 Eric A. Posner and E. Glen Weyl, Radical Markets, Uprooting Capitalism and Democracy for a
Just Society, (Princeton University Press, 2018); Imanol Arrieta Ibarra et al., ‘Should We Treat
Data as Labor?’, American Economic Association Papers & Proceedings, Vol. 1, No.
1,( December, 2017),
17
4.3.2 Identifying data assets
One benefit of the General Data Protection Regulations (GDPR) has been to give every
incentive to organisations to identify and register all their data assets properly. Once
located, data assets can be graded in terms of: key material that needs to be carefully
managed and fully secured (the corporate crown jewels); information that is useful, but
non-personal, requiring basic security; and data that there is no need to keep, and
might indeed become a liability, which needs secure disposal.
4.3.3 Personal or non-personal data?
It would be helpful to establish a clear distinction between personal data, and that
which has no link to specific individuals. The exceptional value, for example of data
related to in-service performance of engineering systems, has been highlighted in
numerous case studies. 13 There is a strong case for the additional value that can be
extracted by controlled and secure sharing or trading of such data. Other countries
may already have a lead on the UK. 14 Such trading must retain secure ownership
through watermarking, audit trails and demonstrably enforceable sanctions against
unauthorised proliferation.
Personal data is far more challenging. Many would argue that they should own the
data related to their own commercial transactions, yet in many cases they have little
choice but to relinquish that ownership in order to use commercial platforms. In social
media, the controls to limit personal data access and exploitation can be challenging to
use or even to find… Similarly, citizens are often asked to release more personal data
than is strictly required for example to establish their entitlement to access particular
services. A market-led response could be through the broad use of trusted
intermediaries 15 to manage and accumulate personal data on customers’ behalf and
perhaps to return some economic value to them?
Personal health data is perhaps the key test case. Many would be content to release
their personal health data for genuine medical research purposes, 16 but would demand
that it not find its way into the hands of, say, life insurance companies… Transparency
and clear authorisation (e.g. opt in rather than opt out) should establish that essential
trust. Innovative solutions, such as that proposed by Sensyne, where NHS Trusts retain
the data sets, with strictly controlled access to data for research purposes, represent a
possible way forward. A charge for access to this highly valuable asset could be used to
contribute revenue to the hard-pressed Trusts and this should be explored further…
4.3.4 Data curation
At this early stage of our data-enabled economy, it is difficult to predict how non-
personal (and subject to strict controls, personal) datasets collected today might be
used and cross linked in the future to unlock new sources of value. This implies the
creation of a new profession in data curation – 21st century librarianship? That
profession would ensure that the necessary metadata is stored alongside the data set
itself and that key parameters such as provenance, quality (including measurement
uncertainty, data calibration and error bars), timeliness, repeatability, and so on, are
13 See for example Royal Academy of Engineering Report Connecting Data: Driving Productivity
& Innovation, Nov 2015.
14 See the German Industrial Data Space, Fraunhofer,
https://www.fraunhofer.de/content/dam/zv/en/fields-of-research/industrial-data-space/whitepaper-
industrial-data-space-eng.pdf
15 See companies such as Mydex.
16 For example, the One Hundred Thousand Genomes project.
18
properly recorded. Without such, the stored datasets are essentially valueless. The
establishment of standards in all these areas is an essential prerequisite…
4.3.5 Making the intangible tangible?
More than two hundred years of experience from the start of the Industrial Revolution
have given us extensive accounting and business tools to identify, manage and control
tangible assets. Yet increasingly corporate investment has switched to intangibles,
including software and datasets. We do not yet have the equivalent tools to value and
manage and fully exploit these intangibles and their absence distorts company
reporting and even national assessments of the balance of payments. For example, it
has been suggested that the USA balance of payments deficit would be halved if
statisticians were better able to capture the value of the software developed in Silicon
Valley. 17
4.3.6 Distorting the development of Deep Learning?
The remarkable advances made in the last ten years in Deep Learning to facilitate
highly targeted Artificial Intelligence (AI) applications have been driven by access to
increasingly large datasets as well as commoditised computing. It is broadly
recognised that such AI applications replicate any in-built biases in the training
datasets used. How can more representative datasets be assembled? How can new
start-ups gain fair access to such data? What impact is restricted access to such data
having on fair competition? What is the role of Government in continuing to give open
access to public data sets, and linked, anonymised, non-personal datasets, for example
through extending the role of the Office for National Statistics?
Much has been achieved, but so much still to do…
4.4 The HAT data ownership model: first party IPR for
individuals
Professor Irene C L Ng
Funded through more than £3m RCUK/EPSRC grants, the HAT (Hub-of-All-Things) and
its related projects set out to design and engineer a legal, economic and technological
artefact (the HAT Microserver) capable of storing, processing, transforming and
exchanging personal data and that also assign a set of rights to the data to individuals
themselves. Its objective is that the personal data sitting within the HAT Microserver
can define, sui generis, a new asset class of PPD i.e. person-controlled personal data,
the personal data where intellectual property rights and excludability of the data
(control) is with individuals. To create the PPD asset class, and the artefact that
contains it, the HAT was designed, engineered and built around 11 design principles
derived from the economic properties of data as a digital good.
4.4.1 Principle of Co-production Access Rights without lien
Personal data has the axiomatic property of co-production. It is generated through

human activity, but collected through technology owned by a firm. The individual must
therefore own a technology/device (the HAT Microserver) that is able to collect data in
such a way that both the firm and the individual, as co-producers, would have access
rights to it in real time and on demand. The data accessed must be free from lien and
encumbrances and, subject to prevailing data protection laws, allow both parties to
reuse and re-share.
17 Hidden value in phones could ‘cut US trade deficit in half’, Financial Times 17th May 2018.
19
4.4.2 Principle of alienable rights
Privacy, according to many advocates, should be an inalienable right. Yet the challenge
here is not privacy, but that personal data controlled by organisations often cannot
even be sufficiently isolated to assign rights. While data may not be assigned rights,
databases are protected by US copyright law and the EU database directive. Database
rights are specifically coded laws on the copying and dissemination of information in
computer databases. Individuals must have their own database and database rights
within the HAT Microserver thereby granting alienable rights for the personal data
within and individuals can grant these rights to others for a period according to their
own wishes and for every data point in the database. This must be achieved by the
individual executing a set of software code within the HAT Microserver to grant time and
context dependent exchange of data with low effort. For rights to be assigned without
ambiguity, there must also be suitable isolation of each HAT Microserver database from
one another. A system of HAT Microservers must therefore be a distributed system of
individual HAT Microservers owned by individuals themselves and yet fully interoperable
with one another and able to be aggregated for firms to render services in a scalable
manner.
4.4.3 Principle of Non-rivalrous Consumption
Personal data has an economic property of non-rivalry i.e. consumption of data by an

entity does not prevent another entity from consuming it (Shapiro & Varian 1998). This
implies that each co-producer may consume the data in a way that benefits itself as
well as contract with other parties, without denying the other of consuming and
contracting the same. That means API access from data sources into the HAT
Microserver database on demand through HAT “data plugs” must create a copy of the
data generated but changing the data rights once the data enters the HAT database, so
as to ensure each co-producer have a set of independent rights for the data that sits
within their domain.
4.4.4 Principle of Expansibility
Personal data has the economic property of infinitely expansibility (Rayna, 2008) . That
means a firm’s data of a person can be copied to another space with very low marginal
cost of re-production. The co-producers could hold the same copy of that data in the
same instant that it is generated in their respective technological domains/devices and
have the ability to contract with third parties to continue expanding its use. The HAT
schema (data structure) allow infinite combinations of data values across datasets to
be exchanged as a data product e.g. Tweets only in Boston, locations between 7-9am.
Each of these data values and bundles can be named and then exchanged/contracted
through standard APIs using standard Internet protocols and encryption in real time. In
a similar way, the firm can do the same with their data (subject to prevailing laws on
personal data sharing)
4.4.5 Principle of Excludability
Personal data have an economic non-excludability property, implying that it is near

impossible to exclude others from consuming the data unless there is a legal (e.g.
contract) or technological (e.g. encryption) framework. Excludability of personal data
controlled by individuals must be based on a data contract and/or technological
instrument whereby individuals are in a position to grant and/or deny rights over
personal data usage. HAT Microservers create data debit contracts when granting
rights of HAT data to others and data in transit is SSL encrypted from end to end, in a
similar manner to emails.
20
4.4.6 Principle of Data Derivatives
Personal data have an economic property of recombinant and divisibility (Quah, 2003).
Personal data e.g. location, combined with time e.g. 7-9am, can create a secondary,
derived data product e.g. commuting journey. A new economic good can be construed
as being created when different types of data are combined in such a way that can be
exchanged, which means that combining personal data for new exchanges increases
the underlying asset value of the database. An individual must control the permission
and process of data being combined and transformed (even if it takes seconds) so that
the database value increases. The individual must also control the usage of private AI
tools on the HAT Microserver that creates new data.
4.4.7 Principle of Data as Store of value
Personal data use contracts cannot specify all states of nature nor all future actions
and use of the data, in advance. When there are states or actions that cannot be
verified ex post by third parties, they are therefore not possible to be contractible ex
ante. The literature on incomplete contracts (see Grossman and Hart 1986; Hart and
Moore 1990; Aghion and Bolton 1992; Dewatripont and Tirole 1994) have shown that
the allocation of power matters when it is not possible to specify in advance precisely
how that power should be exercised. Since the value, worth and use of the data is not
known, the power to decide on future uncertain contracts must be in the hands of the
individual. Therefore, the HAT Microserver has to be the store of value for the individual
before a context emerges for an exchange to occur for personalisation or
recommendation of products and a data contract emerges. If personal data is available
in real time and on demand, every data contract will then be complete for a specific
use with no ambiguity and firms have less need to hoard data.
4.4.8 Principle of Data as Medium of Exchange
The value of some personal data can expire (perish) if not used e.g. the need for Hotel
recommendations. It is therefore context and time dependent. Personal data must
therefore be available on demand and in real time to be a superior asset class and to
be an effective medium of exchange for data contracts for personalisation and
recommendation. By way of the HAT Microserver being both store of value and HAT APIs
being the the vehicle for exchange, HAT data, in its standardised form, should be
treated as currency (like GBP, USD). The only missing element is its ability to be unitised
but that can be derived empirically through increase usage, and scale.
4.4.9 Principle of Transparency
The way personal data is stored, exchanged and processed and the way it stays at rest,
in transit and used must be clear and transparently available for scrutiny. The HAT
Microserver must be an open sourced technology, even if services built on it can be
commercial. The processing of data within the HAT must be based on code that is open
sourced and/or standard Internet technologies. The granting of data rights (usage,
exclusion and alienability) must be transparent.
4.4.10 Principle of Trust Anchoring
Trust is a prerequisite of contracts (Göran and Hägg, 1994). While the HAT Microserver
technology has been legally, economically and technically engineered to endow IPR of
personal data to individuals, it still needs to be issued like a private data account,
much like banks issuing savings or current accounts. For the market to form, HAT
Microservers must still be provisioned on license by a trust anchor, which could be the
data brokers or data trusts, as long as there are guarantees either by the state or
through market incentives, to stay trustworthy. The difference is that, with IPR resting
21
on individuals, the transferability of data rights can be achieved through a direct and
complete contract, much like currency payments, even if it is enabled by data brokers
as trust anchors. This would therefore ensure the market viability of data brokers as a
service for individuals. Trust anchors could also create additional middleware services
or governance mechanisms e.g. hierarchical or nested relationships between HAT
owners e.g. parent and child; a power of attorney situation; or create better heuristics of
data sharing practices across apps within the trust anchor’s ecosystem.
4.4.11 Principle of Market Design
With HAT data having a set of transferable rights, it is now a formal economic good that
is possible to create a thin crossing point (Baldwin, 2007) i.e. a transaction boundary
for the transfer of rights. Matching of HAT data to apps should be dictated by market
design rules of thickness, reduced congestion and safety (Niederle et al., 2008). Best
practices of data exchange should be made transparent and allow different types of
apps (and different levels of exposures) to play out that will optimise choice and
privacy/security concerns.
4.4.12 Implementation
The HAT proof of concept was implemented in November 2016 on AWS (Amazon cloud
service) as the first installation and the ability to generate a HAT Microserver (complete
with a database) within 3 seconds of signing up was achieved in July 2017. The
implementation of the HAT Microserver was optimised to test its cost structure and a
cost of £2 to £4 per month was achieved in January 2018. The HAT is now in live use
both in the innovation environment and in live commercial environment . HATs are
open sourced under AGPL, portable and can be issued from most devices e.g. HATs in
the cloud by different cloud operators; HATs on a Raspberry Pi or even on a PC at home,
or in other devices. However, the security architecture and threat models would differ
for each installation, as would be the business models. While one person per HAT
would dis-incentivise hacking (a hack of one yields one HAT’s data), more work could be
done from the security perspective for different type of HAT installations.
4.4.13 Conclusion
The HAT Project’s ultimate objective is that an explicit, primary market for personal
data, similar to the emergence of a primary market for digital music in the early 2000s,
would reduce illegal and inefficient personal data markets as well as reduce
externalities relating to privacy, as future applications switch to using HATs as user
accounts. The HAT model sets up a parallel asset class to challenge the OPD asset
class through easier access, higher quality and lower friction, much like the way music
licensees challenged music piracy. The HAT full technical system architecture can be
seen at https://developers.hubofallthings.com and the ecosystem at
https://www.hubofallthings.com/the-hat-ecosystem/. A simplified explanation of the
HAT is available at https://www.hubofallthings.com/main/what-is-the-hat/. To date,
there are 1500 HAT owners and the platform on AWS is live. Individuals can obtain a
HAT at the applications live on HATStore https://HATDeX.org/hatstore. To date, there
are 1500 HAT owners and the platform on Amazon Web Service is live, with 12 pilots
that are in various stages of integration with HATs.
This paper is an excerpt submitted to the seminar on the 3 October on Data

Governance with the British Academy, Royal Society and TechUK.
The full paper can be viewed at:
22
Ng, Irene C.L. (2018), “Can you own your personal data? The HAT Data Ownership
Model”, University of Warwick Service Systems Research Group Working Paper series,
ISSN 2049-4297 no. 03/18, at http://wrap.warwick.ac.uk/108357/
23
4.5 Data ownership and data rights

Roger Taylor, Chair of the UK Centre for Data Ethics and
Innovation
Artificial intelligence and the algorithms that determine our experiences both offline
and online are having a profound, and sometimes unexpected, impact on our lives and
society. Questions about what rights and controls individuals, communities and
organisations should have over data sit at the heart of how to unlock the benefits of
these data-driven technologies.
How we answer these questions matters for what kind of society we want to be: who
has the right to what data and for what purpose can support or undermine power
dynamics. It matters for our economy: how we assign rights and values to data can help
stimulate data trade and transfer and encourage innovation. It can also give rise to new
business models and innovation that offer the individuals more control over data about
them.
How we talk about data rights and control also matters. Relying on overly simple
concepts of ownership risks shaping the narrative in unhelpful ways. Perhaps the only
point of agreement at the recent Seminar on Data Ownership, rights and controls:
reaching a common understanding was that traditional notions of ownership are going
to be inadequate. That in itself shows the scale of the challenge we face.
There are two major drivers highlighting why this is a challenge we need urgently to
address.
Firstly, there is a sense that there is a growing public concern about how data about
individuals is used. This is fuelled by scandals such as the much reported-on
Cambridge Analytica, by an increasing awareness of unfair practices such as price
discrimination, and by a worry that a new digital society might lead to the exclusion of
some groups. At the same time, there is no clear alternative emerging from calls for
greater individual control nor any agreement on what this might actually mean. Key
elements have been identified – greater public involvement in determining what is
acceptable within society, greater individual control over how data is used, better
auditing and monitoring of how systems are behaving. But there is much work to do to
define scalable and enforceable solutions.
Secondly, there is a looming threat of missed opportunities unless we have the right
mechanisms to enable the trade and transfer of data. For the UK to continue to be a
world-leading digital economy, we need to have the right legal and governance
frameworks to enable ethical innovation. This means questioning whether some of the
frameworks we have relied on in the past are the right ones to create the best possible
future.
The Centre for Data Ethics and Innovation will play an important part in how we go
about tackling some of the challenges that lie ahead. Our high calibre Board was
appointed in November and we have been tasked by the UK Government to maximise
the benefits of data and AI for our society and the economy.
The Centre will work collaboratively to strengthen the data ecosystem by addressing
public trust and ensuring governance is effective, enabling ethical innovation to
flourish. It will do this by working across sectors providing evidence based advice and
recommendations to policy makers. Our work will include analysing the landscape to
map emerging harms and opportunities. We will also undertake specific projects - our
first, algorithmic bias and online targeting, were announced in the recent Budget. As an
independent body, our advice to Government will be robust, evidence based and reflect
24
engagement with a wide-range of stakeholders including academics, business and the

public.
25
4.6 Reflections on ‘data ownership’

Guy Cohen, Strategy and Policy Lead, Privitar
The arguments against data ownership are numerous and convincing. For one, data
often doesn’t just relate to one person, so to whom does it belong in the first place? If
person a buys something made by person b from person c in a shop owned by person
d, who owns the data of that sale? Data is often about our interactions, and as such
can’t easily be attributed to just one person. Even something as intensely personal as a
date of birth is also information about someone else; the date their mother gave birth.
Second, it is important that individuals have rights over data about them, and that
these rights cannot be sold. So, assuming that our rights would persist, what would it
mean to own someone else’s data? If it is simply giving access to someone who would
otherwise not have access, then what does ownership add over contract law?
So if ownership is not a helpful concept, why is it getting so much attention and

support? I believe data ownership is responding to two related but distinct concerns.
The first is about control. People feel that they are not in control of how their data is
being used, and are concerned that they may be at risk or disadvantaged in some way,
they experience some new uses as ‘creepy’. Second, people feel that companies are
deriving enormous economic benefit from using their data, but they don’t receive any
share of this benefit, and they think this is unfair. Seeing it as two separate issues; one
on data protection rights, and one on economics, allows us to look more specifically at
what we hope to achieve, and thereby identify more appropriate responses.
The first issue, around control, already has a response from data protection law. When
outlining the purpose of the GDPR, Recitals 6 and 7 state that:
“Rapid technological developments and globalisation have brought new challenges for
the protection of personal data. The scale of the collection and sharing of personal
data has increased significantly... Those developments require a strong and more
coherent data protection framework in the Union…Natural persons should have control
of their own personal data. Legal and practical certainty for natural persons, economic
operators and public authorities should be enhanced.”
The question of rebalancing control is one the GDPR aims to answer directly by
strengthening individual rights, enforcement powers, and controller obligations.
Whether the new law will be successful in this goal remains to be seen. But, given it is
still so young, we must wait and see if the GDPR will sufficiently rebalance control in
favour of the individual. The second question, around sharing in the economic benefit
of the big data age, is not really addressed by the GDPR, and I think does require
greater attention.
That the big data age is increasing inequality by reducing the role of labour and
increasing the share of wealth going to capital is a well discussed topic. For data
ownership to lead to the redistribution of wealth a mechanism would be needed for
individuals to be paid for use of their data. This system was advanced by Jaron Lanier in
his 2013 book ‘Who owns the future’, where he suggested a system of micro payments
for data use. Whilst Lanier provides an insightful critique of how the concentration of
wealth occurs, solutions of this kind face many challenges. For instance, the
relationship between the value created by a digital product and the data contributed by
a given individual is not clear. Similarly, a payments system for the amounts likely to be
generated could be extremely expensive, possibly costing more to be managed than the
amounts being distributed. But, more fundamentally, these approaches may have the
undesired effect of inhibiting data use and reducing innovation.
26
The popularity of welfare capitalism in the twentieth century over other models, such as
socialism or libertarianism, is arguably because it has led to better standards of living
for more people. History’s example is that layering protections on top of productive
systems (taxing income to provide welfare), as opposed to taking steps which may
inhibit their productivity (such as socialism), arguably leads to better overall results. For
the data age, valuable productivity comes from the wide use of data, and so steps
which make it harder to use data should be challenged, their impact appraised, and
their necessity demonstrated.
Aside from being problematic as a solution, focusing just on the issue of redistribution
of wealth runs the risk of being blinkered and not looking at how one response might
affect the wider system. Data is often described as the fuel for the current industrial
revolution. Policies which change the way in which this fuel can be used and consumed
risk massively affecting other elements of that system, perhaps positively, but perhaps
not. As such they shouldn’t be considered in isolation, but as part of a systems thinking
approach. Inequality is certainly not the only concern exacerbated by the big data age;
cascading risks and antitrust issues are two other examples that share a common
cause.
Data ownership may be an appealing response to pressing concerns, but it is the wrong
answer to the underlying issues. To better understand what the right answer is, we
need to look more closely at what these concerns are, and look at them in the context
of the data driven industrial revolution and the wider set of associated issues. Policy
makers may need to act, but in doing so they should think in terms of how their policies
will affect the whole system, rather than looking for point solutions for individual issues,
and, crucially, seek solutions which do not put the huge promise of the data age at risk.
Current proposals for data ownership fall short of these objectives.
4.7 Reflections on the data ownership, rights and controls

seminar from the ICO
Romin Partovia, Senior Technology Officer, Information
Commissioner’s Office
On 3 October 2018, the Information Commissioner’s Office (ICO) was invited to the
British Academy, Royal Society and techUK seminar to discuss data ownership, rights
and controls. It was a great opportunity for the ICO to get some feedback from
businesses within the technology sector on their thoughts around data ownership,
building relationships with customers and how the General Data Protection Regulation
(GDPR) fits into this.
Within the GDPR data ownership isn’t explicitly defined as its own concept. We have
data controllers, the organisations that determine the purpose and means of
processing personal data, and data processors, the organisations that can be used to
process personal data on behalf of the controller. The question discussed at the
seminar was who owns this data.
This isn’t necessarily the first question we ask. Primarily we want to know who is in
control of the data – that is who has overall responsibility for managing the data.
Ultimately, this is the data controller, and the GDPR sets out certain responsibilities
that the controller must uphold. This all sounds relatively simple, but often there can be
complex layers of controllers and processors involved and identifying who has what
responsibility can be no mean feat. Our Data Protection Act 2018 guidance on
processors and controllers is a good starting point in establishing if you are a controller
or processor. Moving on, our GDPR guidance on contract and liabilities between
27
controllers and processors will help you manage the relationship and identify your
responsibilities. These can be found on our website:
• Data controllers and data processors: what the difference is and what the
governance implications are
• ICO GDPR guidance: Contracts and liabilities between controllers and
processors
I haven’t mentioned the most important person yet, the person who the data identifies.
We refer to this person as the data subject. Does the data subject own the data that
they give to controllers? They certainly have rights over the data, and these are set out
in the GDPR. Right of access, rectification, erasure, portability and the right to objection
are all articles within the GDPR that empower the data subject. These rights aren’t
always absolute, so controllers should be aware of what rights the data subjects have
and when they can be exercised.
Rather than trying to define data ownership as a legal concept, perhaps it should be
viewed more as a philosophy for processing personal data. If an organisation instils a
culture within the organisation where data belongs to a person and that person owns
the data, it’s a good starting point for building better relationships with customers. This
won’t fit all organisations, so you should be careful in giving data subject’s false
expectations.
The GDPR can be used as a great tool for building trust, retaining and attracting
customers and gaining competitive advantage. In the digital age that we live in,
organisations should be striving for privacy and innovation; the GDPR allows us to
achieve this.
Data protection by design and default provide the opportunity to include data protection
practices into your processing activities and business practices. Data Protection Impact
Assessments (DPIAs) allow you to identify and minimise data protection risks of a
project, and Article 32 of the GDPR sets out the security requirements of controllers, so
customers trust their data will be safe in your hands. These are just a few examples of
how the GDPR can help build better relationships with customers. The organisations
that embrace these will be the organisations that have better relationships with
customers.
We had great questions from the audience about anonymisation and the increase in
data protection breaches.
There was concern that organisations could sell or buy personal data that was
anonymised first, especially with the increase of data-hungry Artificial Intelligence
systems. Once personal data is anonymised it no longer becomes personal data and
GDPR requirements no longer exist. However, with that said, truly anonymised data is
difficult to achieve, and it is not to be confused with pseudonymised data.
Pseudonymised data is still personal data and will require compliance with the GDPR.
Where ever that personal data goes, the rights of the data subject go with it. Further
guidance can be found on the ICO website.
A great point was made about the increase in personal data breaches and if this will
become an upwards trajectory of ever-increasing breaches. The ICO has certainly seen
an increase in calls to our helpline. We see this as a sign that people are becoming
more aware of their data protection rights and organisations are coming to us for
advice. The ICO has also been strengthened in number and expertise, allowing us to
carry out our responsibilities and obligations to organisations and to the public.
Whereas data protection by design was once a best practice, the GDPR now requires it
as a requirement of processing personal data. This will ensure the ICO can take
proactive steps in ensuring data minimisation and other good practices are adopted.
28
We are in the process of setting up a regulatory sandbox that will support organisations
to develop innovative products and services using personal data in different ways. We
also have a grants programme to support innovative research and solutions focused on
privacy and data protection issues.
The GDPR now requires organisations to carry out DPIAs before processing data that
are likely to result in high risk to an individual’s interest. We have a DPIA team that can
support you and provide feedback on your DPIAs to ensure good data protection
practices.
These are all tools that the ICO is championing to ensure that organisations can follow
good data protection practices, whilst also improving an organisation’s ability to retain
and attract customers, innovate new products and services and develop new ways of
thinking.
4.8 Data ownership, rights and control – an anthropological

perspective
Dr Hannah Knox, Associate Professor of Anthropology,
University College London
This paper is the script of the presentation that Dr Knox gave at the seminar.
I have been asked to comment on some alternative ways in which we might think about
data ownership. As an anthropologist I’m lucky to be part of a discipline for whom the
question of what it means to own something has been very central. Given this I want to
try and convey some of the ways in which anthropologists have thought about
ownership and to consider what the relevance of these approaches might be for the
questions that we are trying to tackle here today about the ownership and use of data.
No discussion about ownership in anthropology would be complete without locating it in

relation to the social practice of exchange. For it is primarily at the moment of exchange
– when one person gives something to another person, when the question of
ownership is made most visible. To understand what it means to own something,
means understanding the conditions under which that thing can be transferred to
somebody else.
One of the things that anthropologists have repeatedly observed, is that to exchange
something is an act that requires that thing become detached from the person who
previously owned it – socially, materially, legally - and to pass it on to another person or
body who becomes newly attached in some way to that thing (Weiner 1992). But not all
things do this act of attachment and detachment in the same way. Recognising this, a
key distinction that has emerged in anthropological studies of exchange, has been the
difference between gifts and commodities (Gregory 1982).
Perhaps then, what is at stake in debates about data ownership and use, is not only
data’s legal status as property, but also its social status as an artefact of exchange.
Could problems about what constitutes appropriate exchange in fact be what is at the
heart of discussions about data ownership?
To consider this I want to use the example, discussed by anthropologist Marilyn

Strathern, of a prior debate about public policy and ownership that has many parallels
with the data debates we are having here today – the practice of organ donation
(Strathern 2012). Like data ownership this has been a fraught area with many of the
same ethical arguments about appropriateness of exchange being rehearsed.
29
Strathern points out that debates about the donation of organs have frequently hinged
on whether the act of giving an organ should ideally be an altruistic act – that is, a gift -
or whether people should be monetarily compensated for their body parts. For those
who advocate a non-monetised form of organ donation, the act of exchange is one of
giving a gift. The language of altruism implies that the gift is entirely disinterested. It
also implies that after having given the gift there need be no reciprocal return. However,
that is not the end of the story. Those who have received an organ often want to repay
the donor, meanwhile the donor themselves often articulate their act not as simply
altruistic but rather as a public act for the greater good.
At the same time the gift of giving introduces some other tensions. For example, the
idea of giving away a body part as an act of altruism - and the language that is used to
describe this act as one of gifting – also hides another feature of organ donation which
is often commented on – that ‘human donations enter the organ procurement and
distribution system altruistically, and exit commercially’ (Strathern 2012, 405). We
might argue then, that language of the gift here does important work of concealing a
commoditisation process that is at play in the world of organ donation. A recognition of
this tension, has opened up discussions about whether in fact, the commodity relation
– i.e. the monetisation of organs within the healthcare system – should be brought
back into the moment of organ of donation. If someone is creating monetary value out
of an organ, then should the act of giving not also be adequately compensated?
Otherwise, the worry is that there is an unevenness in the exchange – a one-sided gift
which now appears to deserve a reciprocity, but which does not receive it.
A similar dynamic can be seen at play in discussions about data. The way in which we
think about data from an individual point of view often uses the language of gift
exchange as well. People are asked to ‘give’ their consent for data to be used, or we
talk about sharing our data with others. When individuals articulate that current
regimes of data exchange are unproblematic for them, they frequently invoke a
language of reciprocity to resolve the movement that makes their data someone else’s
(for example people say that they recognise that if they don't give their data they won’t
have access to services like Google Maps, so there is a reasonable exchange at play
here). Conversely for those who see data exchange as problematic, it is precisely the
mismatch between the free gift of data on the one hand and the way in which that data
is monetised on the other, that is at stake. The surprise that is often expressed when
data is revealed as a source of revenue generation derives from the same disjuncture
between the language of gifting data at the moment of its generation and the
commercialisation of data as it is monetised and exchanged that we saw in organ
donation. Following from this come ideas such as giving individuals micropayments for
their data to balance the seemingly uneven nature of exchange.
There is much more we could say on this. For example, when data is co-produced by
corporations and so called ‘prosumers’ what effects does this have on the way in which
people negotiate data exchange? Another question concerns personal data and the
relationship that data has to the body or self. How do people let go of things that are
conceived as still being a part of themselves? And what are the social implications
when personal data returns as a ‘digital double’ (Knox et al. 2010)? Here the question
become less one of who owns data, and more one of how to successfully achieve
exchange, when gifts turn out to be commodities and data-objects that were given away
can never be fully separated from their previous owners.
References
Gregory, C. A. 1982. Gifts and commodities, Studies in political economy. London ; New
York: Academic Press.
30
Knox, H., D. O'Doherty, T. Vurdubakis, and C. Westrup. 2010. The Devil and Customer
Relationship Management. Journal of Cultural Economy 3 (3):339-359.
Strathern, Marilyn. 2012. Gifts money cannot buy. Social Anthropology 20 (4):397-410.
Weiner, Annette B. 1992. Inalienable possessions : the paradox of keeping-while-giving.

Berkeley ; Oxford: University of California Press.
31
5.0 List of speakers

Co-chairs
Prof Genevra Richardson Professor of Law, King’s College London; Chair of the
CBE FBA British Academy Public Policy Committee
Prof Dame Otteline Leyser Professor of Plant Development, University of
DBE FRS Cambridge; Chair of the Royal Society Science Policy
Advisory Group
Sue Daley Head of Programme Cloud, Data Analytics & AI,
techUK
Keynote speakers
Prof Sarah Worthington Downing Professor of Laws of England and Director of
FBA the Cambridge Private Law Centre, University of
Cambridge
Roger Taylor Chair, Centre for Data Ethics and Innovation
Panel 1: Exploring the concept of ownership, its value and limitations

Benoit Van Asbroeck Partner, Bird & Bird
Rachel Coldicutt CEO, doteveryone
Dr Hannah Knox Associate Professor of Anthropology, University College
London
Jeni Tennison CEO, Open Data Institute (ODI)
Panel 2: Building data rights and relationships between consumers, businesses and
the public sector
Simon Burall Senior Associate, Involve
Emma Butler Data Protection Officer, Yoti
Jacqueline Davey Vice-President Sales in the UK and Ireland, IBM
Romin Partovnia Senior Technology Officer, ICO
Panel 3: Reflections
Sophia Adams-Bhatti Director of Legal and Regulatory Policy, Law Society
Guy Cohen Strategy and Policy Lead, Privitar
Hetan Shah Executive Director, Royal Statistical Society
Prof Karen Yeung Interdisciplinary Professorial Fellow in Law, Ethics and
Informatics, Birmingham Law School, University of
Birmingham
32
The British Academy is the UK’s national body for the
humanities and social sciences – the study of peoples,
cultures and societies, past, present and future.
www.thebritishacademy.ac.uk/ | @ @BritishAcademy_
The Royal Society is a Fellowship of many of the world’s most

eminent scientists and is the oldest scientific academy in
continuous existence.
royalsociety.org/ | @ royalsociety
techUK represents the companies and technologies that are

defining today the world that we will live in tomorrow. Over
900 companies are members of techUK.
techUK.org | @techUK

Data Ownership Rights and Controls October 2018

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Data Ownership Rights and Controls October 2018

Uploaded by

Copyright:

Available Formats

Data ownership,

rights and controls:

3.0 Data ownership, rights and controls: discussions at a

3.1 Exploring the concepts 4

4.0 Contributor papers 10

4.1 Legal notions of ‘property’ and ‘ownership’ -

5.0 List of speakers 28

3.0 Data ownership, rights

This sense of unease is a difficult problem to pin down, especially as it is in tension

3.1 Exploring the concepts

3.1.1 Data ownership

1 , Including the Facebook-Cambridge Analytica scandal in March 2018.

Anthropology considers ownership in relation to the social practice of exchange. It is

3.1.2 Consent and control

3.1.3 Data rights

Additionally, a rights-based approach helps to ensure at least a minimum degree of

However, existing conceptions of human rights are highly individualised in their

3.2 Moving the debate forward

3.2.1 Taking a societal view

3.2.2 Understanding the value of data

3.2.3 Focus on use

3.3 Getting to where we want to be

An approach that focuses on building a trustworthy system and providing stewardship

3.3.2 Engaging the public and building a social vision

We do not have a sufficiently deep understanding of public attitudes with regard to

4.0 Contributor papers

4.1 Legal notions of ‘property’ and ‘ownership’

4.1.1 What is the regulatory end goal?

4.1.2 ‘Property’ is not as protective as one might think

4.1.3 How do lawyers think about ‘property’?

Notice that such an arrangement is necessarily a transactional arrangement between

4.1.4 Data as property?

4.1.5 Alternative strategies beyond ‘data ownership’

Most legal protections of individuals can be waived by consent. Increasingly we give

4.2 Data ownership: is it an appropriate concept?

People tend to assume an implicit understanding of the concept of data ownership,

Data is produced through transactional relationships so it is difficult to ascribe it

effectively in the public sector, (Reform, 2018).

Session 2017–19 (House of Lords, 2018), 28.

Review, 27, 2001, 142.

4.3 Data: vital asset or toxic liability?

4.3.2 Identifying data assets

4.3.3 Personal or non-personal data?

4.3.4 Data curation

16 For example, the One Hundred Thousand Genomes project.

4.3.5 Making the intangible tangible?

4.3.6 Distorting the development of Deep Learning?

Much has been achieved, but so much still to do…

4.4.1 Principle of Co-production Access Rights without lien

Personal data has the axiomatic property of co-production. It is generated through

4.4.2 Principle of alienable rights

4.4.3 Principle of Non-rivalrous Consumption

Personal data has an economic property of non-rivalry i.e. consumption of data by an

4.4.4 Principle of Expansibility

4.4.5 Principle of Excludability

Personal data have an economic non-excludability property, implying that it is near

4.4.6 Principle of Data Derivatives

4.4.7 Principle of Data as Store of value

4.4.8 Principle of Data as Medium of Exchange

4.4.9 Principle of Transparency

4.4.10 Principle of Trust Anchoring

4.4.11 Principle of Market Design

This paper is an excerpt submitted to the seminar on the 3 October on Data